The future of AMD in graphics

[tamz_msc]

There is the 25x20 efficiency initiative in AMD for both CPUs and GPUs.

That is a pretty silly metric as it's performance/ idle power. It's all PR as the real world power efficiency of RR is piss poor.

 

[AtenRa]

That is a pretty silly metric as it's performance/ idle power. It's all PR as the real world power efficiency of RR is piss poor.

This is not perf/idle , its Cinebench / 3D Mark 11 divided by energy usage .

 

[tamz_msc]

This is not perf/idle , its Cinebench / 3D Mark 11 divided by energy usage .

Energy usage in that metric is defined by a composite of different kinds of idle power.

 

[NTMBK]

If they won the next gen consoles, I think they'll be fine. Hopefully for Navi they cooked up some cool new features that all the next gen games use.

EDIT: I remember before the PS4 was formally announced, lots of people were arguing that AMD couldn't possibly have got the console contracts, because who would want to tie themselves to a company that was obviously doomed? They did fine then with a weak CPU, and they'll do fine now with a weak GPU.

 

[exquisitechar]

AMD and Intel will likely go nowhere in the Datacenter GPU space unless Nvidia falls over flat on its face. Past execution suggests Nvidia typically outmaneuvers both, because its a better run company. Oh, and Nvidia is the incumbent with CUDA.

Not sure about AMD but I don't think AI/ML will end up much for Intel. They tend to pursue side projects they abandon in 5-7 years at most. The projects usually start because they are too late and are following a trend.

Big companies like Intel have too much resources and don't know what to do with it outside their core competence. Too much veering from it, and the efforts are tossed in the garbage because its financially motivated, not for technological advancement or being a pioneer at it.

I was just thinking the opposite. We're already seeing AMD taking market share in datacenter GPUs with their weak Vega offerings (data center revenue split between CPU and GPU was surprisingly stated to be roughly similar during AMD's last ER), supposedly by leveraging the early enthusiasm in the industry for EPYC Rome. When Intel makes their incursion into this market and likely does the same thing, making full use of being a provider of both CPUs and GPUs, I think Nvidia will be in for a world of hurt.

Similarly, I think Intel could dominate the low to mid range with their long delayed updated GPU uarch on 10nm. I can especially see them doing well in laptops with the technology that they have recently talked about and their clout with OEMs makes them a serious danger to Nvidia as well, whose bread and butter are clearly still desktop/laptop dGPUs as their recent earnings reports indicate, with the evidently catastrophic effects of crypto and weak RTX 2000 series sales (despite their tendency to place the emphasis on things like their automotive efforts).

For AMD's dGPU efforts, Intel's GPUs are bad news too. There may be no room for a third player in this market, and if I had to predict who will be in the weakest position, it would be them. If they don't make massive improvements to their GPUs with Navi and Navi+1 they can forget about this market, and even if they will, as always, Nvidia will probably outsell them even with similarly performing but much higher priced GPUs, not to speak of Intel taking their share as well. Their staggered release schedule for APUs also makes them non competitive in this segment, with Intel updating both their CPU uarch and becoming much more competitive on the GPU front with Gen11. I think they squandered the advantage of offering both strong CPUs and GPUs when it comes to APUs, now that Intel is investing heavily in GPUs as well. It is easy to understand why their Ryzen APUs are low priority compared to other segments, but it's still such a waste. Another segment where their graphics could have been much more successful...

At least AMD still has their semicustom division with the next gen MS and Sony consoles almost certainly running on Navi. I think they are also serious about their datacenter GPUs, and Mi-NEXT will probably do decently too.

 

[railven]

Raja focused on playing stupid power games. Obviously RTG was formed due to his idea and his obvious plan was to sell it to intel. AMD GPUs went downhill from the moment he was head and never recovered since. It's a good thing he left for AMD.

Sometimes posts like these remind me how quickly people turn on each other (not you, poster). I remember when Raja was some savior, and the marketing of "being pioneers" were hints of AMD"s incoming revolution with Raja at the helm. Then we got Vega, Raja went on hiatus, but no he wasn't leaving not le king, then he left, and now he's no one of relevance (unless he hits a homerun at Intel, then he's a traitor). My favorite time this happened was while reading Reddit someone posted "Glad Raja is gone, bring back the guy that was around during HD 4K/5K series." Only to be responded with "That was Raja." haha.

Anyways, AMD - I personally see them just sticking to consoles and low/mid tier GPUs. Seeing their performance in the CPU side, I feel like AMD is focusing on regaining marketshare first. They'll do that by chipping away at Intel/Nvidia's volume sales. I don't think AMD is going anywhere, just not moving the needle much for a few more generations. Intel's next move can shake AMD's boat more in my opinion than anything Nvidia does. Only thing I can see Nvidia doing to hurt AMD is stealing console contracts. Happened with Nintendo, who knows what else can happen. /shrug

 

[exquisitechar]

Sometimes posts like these remind me how quickly people turn on each other (not you, poster). I remember when Raja was some savior, and the marketing of "being pioneers" were hints of AMD"s incoming revolution with Raja at the helm. Then we got Vega, Raja went on hiatus, but no he wasn't leaving not le king, then he left, and now he's no one of relevance (unless he hits a homerun at Intel, then he's a traitor). My favorite time this happened was while reading Reddit someone posted "Glad Raja is gone, bring back the guy that was around during HD 4K/5K series." Only to be responded with "That was Raja." haha.

Anyways, AMD - I personally see them just sticking to consoles and low/mid tier GPUs. Seeing their performance in the CPU side, I feel like AMD is focusing on regaining marketshare first. They'll do that by chipping away at Intel/Nvidia's volume sales. I don't think AMD is going anywhere, just not moving the needle much for a few more generations. Intel's next move can shake AMD's boat more in my opinion than anything Nvidia does. Only thing I can see Nvidia doing to hurt AMD is stealing console contracts. Happened with Nintendo, who knows what else can happen. /shrug

I think it's more likely that Intel will steal AMD's console business than Nvidia.

 

[Muhammed]

But AMD actually have the lead over Nvidia, they will release 7nm graphic cards somewhere around June/July and will have the process lead for over 6 months, likely 9 months before Nvidia can come with a new architecture.

Even with that, Radeon VII can't beat a 12nm 2080. And Instinct 60 can't beat a 12nm V100. It's not a lead, it's a curse. When NVIDIA drops 7nm it's going to be a blood bath.

Even Nvidia's Volta is too weak across a wide range of computational workloads. It might beat AMD offerings in few workloads by a big margin, but it losses across a wide range.

Source for that? Becuase this info is total bogus. Even AMD admits to occasionally offer the same performance as V100. And the V100 is selling like hot cake, order of magnitude higher than Instinct.

Furthermore AMD can lower prices and stay competitive that way. I would definitely buy a Vega 56 over the 1070 or 1660ti at the same price. RX 580 8GB has been as low as $200, RX 570 as low as $160, these are tremendous value. GTX 1060 6GB has not moved bellow $250, apart from very few specific sales and promotions or low cost single slot cards from partners like Evga.

As usual continuing to sell based on price which means they are not really selling anything. They are reducing the price to coax some buyers to buy AMD, but most people buy NVIDIA. In the latest Steam Survey, the 1060 outsold both the 480+580 combined by 12X! That's a HUGE difference.

NVIDIA now has the power efficiency advantage, the mindhsare, marketing, IQ advantage (with ray tracing), new technology, and even features. AMD has none, except to reduce prices, like they did with their Bulldozer lineup.

 

[maddie]

Sometimes posts like these remind me how quickly people turn on each other (not you, poster). I remember when Raja was some savior, and the marketing of "being pioneers" were hints of AMD"s incoming revolution with Raja at the helm. Then we got Vega, Raja went on hiatus, but no he wasn't leaving not le king, then he left, and now he's no one of relevance (unless he hits a homerun at Intel, then he's a traitor). My favorite time this happened was while reading Reddit someone posted "Glad Raja is gone, bring back the guy that was around during HD 4K/5K series." Only to be responded with "That was Raja." haha.

Anyways, AMD - I personally see them just sticking to consoles and low/mid tier GPUs. Seeing their performance in the CPU side, I feel like AMD is focusing on regaining marketshare first. They'll do that by chipping away at Intel/Nvidia's volume sales. I don't think AMD is going anywhere, just not moving the needle much for a few more generations. Intel's next move can shake AMD's boat more in my opinion than anything Nvidia does. Only thing I can see Nvidia doing to hurt AMD is stealing console contracts. Happened with Nintendo, who knows what else can happen. /shrug

Taking a composite of various posters and spinning a narrative is very creative. A modern person in all respects. Well done.

I for one defended him at first until I started to see a talented glory seeker with no verbal self control. Anyone remember the PCperspective interview ploy with the various alcohols.

 

[GodisanAtheist]

For years folks were wondering how AMD's debits and deficits were going to affect their product launches. Well, here we are.

I think AMD is going to try and "Zen" their graphics division: process nodes are starting to slow down, the entire industry is starting to slow in terms of cheap and easy improvements, and they're counting on NV slowing it's roll like Intel did.

Then they drop their next 10 year arch and try to claw back as much revenue and market share as possible until their competitors regroup and stomp them through superior R&D budgets, products, mind & market share.

Even the much lauded Zen has merely brought them back into competition with Intel, in terms of performance it's still very much a two horse race. I expect the same thing to happen in GPUs: AMD will release another HD 4xxx/ HD 7xxx series that's not a clear win, but competitive on multiple fronts before dragging the arch out way too long because they just don't have the money (& vision/focus) to constantly itterate on their arch the way NV does.

For all of Dr. Su's talk of AMD is not content being a second fiddle in any market, AMD will likely focus the bulk of their efforts on the sub-$300 market to get the most bang for their buck.

 

[AtenRa]

And Instinct 60 can't beat a 12nm V100. It's not a lead, it's a curse. When NVIDIA drops 7nm it's going to be a blood bath.

MI 60 = 13.23B transistors

V100 = 21.1B transistors

Source for that? Becuase this info is total bogus. Even AMD admits to occasionally offer the same performance as V100. And the V100 is selling like hot cake, order of magnitude higher than Instinct.

I would very much like you to give us those selling numbers you have.

On the other hand we know that AMD Compute GPUs made a record revenue in Q4 2018.

https://seekingalpha.com/article/42...-2018-results-earnings-call-transcript?page=2

We set a record for professional GPU revenue in the quarter, driven by multiple high volume wins for our Vega based data centers GPUs. We started shipping our new 7-nanometer Radeon Instinct accelerators in the quarter and introduced a major set of enhancements to our data center GPU software that make it easier for customers to deploy Radeon GPUs for AI and machine learning workloads.

As usual continuing to sell based on price which means they are not really selling anything. They are reducing the price to coax some buyers to buy AMD, but most people buy NVIDIA. In the latest Steam Survey, the 1060 outsold both the 480+580 combined by 12X! That's a HUGE difference.

Im not saying that GTX1060 would not outsell RX480/580 but those Steam numbers are made simple because of the mining frenzy where gamers only bought NVIDIA cards due to insane AMD prices and low availability.

NVIDIA now has the power efficiency advantage, the mindhsare, marketing, IQ advantage (with ray tracing), new technology, and even features. AMD has none, except to reduce prices, like they did with their Bulldozer lineup.

Yeap, right now AMD wants to clear all RX 5xx and Vega 56/64 stock and then bring Navi to compete against Turing.

[/QUOTE][/QUOTE]

 

[Dribble]

I think AMD's will be working flat out on console gpu's till 2020. They will be a new design, include ray tracing, much better compression, etc. After the consoles are done man power will be available to adapt those designs for pc gpu's and they will finally release something competitive with Nvidia.

Until then on the pc it's likely just going to be more GCN, and they won't really be competitive - I suspect Navi will basically be Polaris @ 7nm.

I have no confidence in Intel to manage to release anything competitive for significantly longer as they have failed in pretty well everything new market they've tried to get into for the last few years. Almost certainly the first thing to come out will be for gpu compute more then pc gaming, the drivers will suck, and it won't really be competitive outside of 1 or 2 theoretical benchmarks. There will be much fanfare, a few big sales (where Intel are basically giving them away) and then nothing.

 

[railven]

I think it's more likely that Intel will steal AMD's console business than Nvidia.

I sort of feel like this is more likely, but at the same time I don't. Intel has their head up too far into their rectum. Unless this new CEO is gonna play some hard ball.

Taking a composite of various posters and spinning a narrative is very creative. A modern person in all respects. Well done.

What narrative? My observations are just that - mine.

I for one defended him at first until I started to see a talented glory seeker with no verbal self control. Anyone remember the PCperspective interview ploy with the various alcohols.

Raja is a great engineer. I'm on the side AMD didn't have the resources. Although with how bad Vega was received, it was probably in AMD and Raja's best interest to part ways. Hopefully he can do more at Intel, because NV needs a strong rival. Desperately.

 

[exquisitechar]

I think AMD's will be working flat out on console gpu's till 2020. They will be a new design, include ray tracing, much better compression, etc. After the consoles are done man power will be available to adapt those designs for pc gpu's and they will finally release something competitive with Nvidia.

Until then on the pc it's likely just going to be more GCN, and they won't really be competitive - I suspect Navi will basically be Polaris @ 7nm.

I have no confidence in Intel to manage to release anything competitive for significantly longer as they have failed in pretty well everything new market they've tried to get into for the last few years. Almost certainly the first thing to come out will be for gpu compute more then pc gaming, the drivers will suck, and it won't really be competitive outside of 1 or 2 theoretical benchmarks. There will be much fanfare, a few big sales (where Intel are basically giving them away) and then nothing.

AMD won't do ray tracing until it can be done across their entire product stack, David Wang said as much. They probably won't do it with fixed function hardware like Nvidia, either. Next gen consoles will be Navi based. I don't see ray tracing being much of a thing in consoles until some mid generation refresh, maybe it will be used very sparingly in certain games.

I think Navi will be plenty competitive and will still be massively outsold by the 1660ti, heh. Navi+1 will be almost surely be GCN too...

 

[IntelUser2000]

I was just thinking the opposite. We're already seeing AMD taking market share in datacenter GPUs with their weak Vega offerings (data center revenue split between CPU and GPU was surprisingly stated to be roughly similar during AMD's last ER),

Why do people have this tendency to think Datacenter is easier than the consumer market? Is it because they understand it less? Nvidia didn't rise to the top there overnight, it took them years and years. Same with Intel in the server space. It took them 10 years before they became to have the monster presence they have now. Both of them won because they consistently delivered outstanding products, offered good support, tools, which all build trust.

Similarly, I think Intel could dominate the low to mid range with their long delayed updated GPU uarch on 10nm.

Man, Intel isn't there to dominate in the low to mid range GPUs. That's a crap business strategy for a company like Intel. They already do good on the low end with existing iGPUs. The only time you'll see them aiming so low is if they can't reach the top performance bracket and they are forced to do so. That's the same as surrendering the market. The money is in the top end.

The last time Intel tried in graphics was the Iris Pro in Haswell. Sure its not discrete graphics, but look at what happened.

Iris Pro 5200 - Apple used it, and one model from a tiny laptop maker(Clevo?)
Iris Pro 6200 - Apple used it, no one else.
Iris Pro 580 - Intel used it on their NUC

They lost because it sucked, period. They had all the disadvantages of being an integrated GPU without the advantages of it. It had battery life as crap as the discrete graphics competition, and the systems were expensive, while the performance was barely competitive with the intended target. They make decent iGPUs that go in their chips. They have not proven they can scale it up and be competitive.

 

[senseamp]

It ultimately comes down to the people. You can't expect massive architectural and design improvements to happen just with passage of time, someone has to come up with them. Right now, the tech world is hot for hardware talent, especially anything GPU or AI ASIC related. Everyone from mobile device makers, to hyperscalers, to startups is building chips now and are raiding talent, dropping tonnes of total comp and/or equity on the kind of people AMD needs to hire/retain to catch up. It's a tough road for a third tier company like AMD. That said, the good news for AMD is you don't fall that far behind by just being mediocre, you have to actively screw up some major things. So while leadership might be a tall order, it's reasonable to at least expect them to close the gap enough to where they can still make a little bit of money off GPUs instead of being boxed into having to price them at or below cost.

 

[exquisitechar]

Why do people have this tendency to think Datacenter is easier than the consumer market? Is it because they understand it less? Nvidia didn't rise to the top there overnight, it took them years and years. Same with Intel in the server space. It took them 10 years before they became to have the monster presence they have now. Both of them won because they consistently delivered outstanding products, offered good support, tools, which all build trust.

It apparently is easier, according to AMD's last ER. With AMD's datacenter dGPU revenue being roughly the same as CPU revenue, it is clear that their revenue share in DC GPUs is not insignificant. In an article on Seeking Alpha, the writer did napkin math that mostly seems to check out, taking datacenter dGPU revenue to be a bit lower than EPYC revenue just to be on the safe side, and got that AMD has around 12.5% revenue share. Just found it, here it is. AMD started from basically nothing in 2017 and, although Vega 20 is decent for certain solutions, the fact is that it is overall not a particularly strong offering. Reaching double digit revenue share in that time is impressive, and it seems to me that the myth of Nvidia having an almost insurmountable position in datacenters is clearly just that, a myth.

Man, Intel isn't there to dominate in the low to mid range GPUs. That's a crap business strategy for a company like Intel. They already do good on the low end with existing iGPUs. The only time you'll see them aiming so low is if they can't reach the top performance bracket and they are forced to do so. That's the same as surrendering the market. The money is in the top end.

The last time Intel tried in graphics was the Iris Pro in Haswell. Sure its not discrete graphics, but look at what happened.

Iris Pro 5200 - Apple used it, and one model from a tiny laptop maker(Clevo?)
Iris Pro 6200 - Apple used it, no one else.
Iris Pro 580 - Intel used it on their NUC

They lost because it sucked, period. They had all the disadvantages of being an integrated GPU without the advantages of it. It had battery life as crap as the discrete graphics competition, and the systems were expensive, while the performance was barely competitive with the intended target. They make decent iGPUs that go in their chips. They have not proven they can scale it up and be competitive.

Currently, Intel's iGPUs only serve to function as the bare minimum in performance. Even if they don't immediately reach the top performance brackets on their this try with dGPUs, they can start leveraging their position with OEMs to put their dGPUs in the majority of laptops which need higher performance than their iGPUs. I think you are too pessimistic about their future competitiveness. We can already see massive improvements from what we know about Gen11.

 

[VirtualLarry]

NVIDIA now has the ... IQ advantage (with ray tracing)

You mean with DLSS? Super-Smeary? LOL.

Edit: Sorry, that was childish. What I was intending to point out, was the irony of someone claiming that RTX has an IQ advantage due to RT, when the only way to run it with any sort of performance for gaming is along with DLSS, which, given some YT reviews (*which may be out of date, NVidia claims to have improved DLSS for some AAA games recently), which accused DLSS of "smearing", and looking worse than just regular AA with rendering resolution scale turned down.

 

[IntelUser2000]

It apparently is easier, according to AMD's last ER. With AMD's datacenter dGPU revenue being roughly the same as CPU revenue, it is clear that their revenue share in DC GPUs is not insignificant.

12.5% share is pretty much what they had with Bulldozer derivatives in the CPU markets. Yea they have presence, but its really not much.

I think you are too pessimistic about their future competitiveness. We can already see massive improvements from what we know about Gen11.

I was not, until they proved many times I was wrong. I believe the GT2 Gen 11 GPU will be on par or slightly outperform Vega graphics in Raven Ridge parts. Some will say I'm being very optimistic there. Actually I'm not, because by the time Gen 11 is here AMD will be here with their next gen. They've been 1 generation behind AMD for quite a long time. And their real competitor is Nvidia, that executes much better.

I think there's potential. However its too early to say they'll be a player in discrete parts any shape or form.

 

[beginner99]

Raja is a great engineer. I'm on the side AMD didn't have the resources

At his position his engineering skill hardly matters as he doesn't do any technical stuff at all. Of course AMD was money-constrained and that is why he was playing power games to get more money or get sold of to intel.

 

[JDG1980]

I think Navi will be plenty competitive and will still be massively outsold by the 1660ti, heh. Navi+1 will be almost surely be GCN too...

The architecture after Navi is supposed to be an all-new architecture. They can't keep iterating on GCN; it's run out of steam and simply is not competitive in the gaming market any longer.

 

[JDG1980]

The rumors/leaks so far indicate that Navi will be a low-range solution. The release of Radeon VII as a consumer product also bolsters this, since it wouldn't make much sense for AMD to put out a card that would be superseded within a couple of months.

As a low-range product, Navi's primary competition will be the new GTX 1660 Ti. In general, AMD has to provide better performance per dollar than Nvidia to gain market share from them. If Navi comes in near the 1660 Ti's performance, then it will need to sell at $249 or less. If it can push a bit further, to approach GTX 1080/RTX 2060 performance levels, then it could sell at the same $279 price point at the 1660 Ti.

Navi will have a process advantage over TU116, but an architectural disadvantage as GCN is clearly not as optimized a gaming architecture as Turing. Assuming they're not going to fix the 4-shader-engine limitation in GCN, the obvious choice for a low-range product would be a core count in the Hawaii range, somewhere between 2560 and 3072 shaders. Currently, the RX 590 offers about 80% of the GTX 1660 Ti's performance at 1080p and 1440p. That means AMD needs to boost performance by about 25 percent over what they're getting out of Polaris to have a viable product in this segment. Polaris has GDDR5 clocked at 8Gbps; GDDR6 can offer speeds of 12-14Gbps. Even the low end of that spectrum would mean a 50% increase in memory bandwidth, assuming that they stick with a 256-bit bus (and they probably will). If they went with a shader count of 2880 (same as Hawaii) that would be a 25% increase over Polaris. How far they can push clocks is an open question; they can probably do better than the Radeon VII's 1750 MHz boost clock, given a lower shader count and perhaps some architectural improvements for higher clock speeds. Overall, a 25% performance boost over Polaris for Navi seems realistic, but whether they can do this in a sane (150W) TDP is another question. RTG doesn't even seem to recognize the existence of performance-per-watt (at least on the consumer side), which is a real problem.

 

[beginner99]

The rumors/leaks so far indicate that Navi will be a low-range solution. The release of Radeon VII as a consumer product also bolsters this, since it wouldn't make much sense for AMD to put out a card that would be superseded within a couple of months.

Well Navi will be late 2019 or early 2020 so the VII has almost a year on the market and it was not a gaming card to begin with or said otherwise also offering it as a gaming card, even if only for 9 months, didn't really cost AMD much. I would expect Navi to come close to VII at much lower power use and much lower price.

 

[Guru]

hey won the next gen consoles, I think they'll be fine. Hopefully for Navi they cooked up some cool new features that all the next gen games use.

Well Navi will be late 2019 or early 2020 so the VII has almost a year on the market and it was not a gaming card to begin with or said otherwise also offering it as a gaming card, even if only for 9 months, didn't really cost AMD much. I would expect Navi to come close to VII at much lower power use and much lower price.

Actually Lisa Su confirmed herself that we are looking at early H2 release on the rest of their GPU lineup. I think their processors Ryzen 3000 will come April through June, then have the first Navi GPU in July.

AMD are going to have a process node advantage for at least 6 months though, so I'm sure they are going to use that. They are also not rushing Navi, they are taking their time with it, and working with Vega to release new products in the professional market and now with Radeon 7 in the gaming market.

I think we might see another Vega 20 offshoot, either with its full shaders or even more cutdown in order to compete at the lower price range.

I think Navi is going to be their all round gaming architecture, from the low end to the high end. I think they are going to maintain the Vega series for the compute/professional market and iterate on it for that market, while working on Navi for the gaming segment over the next 2-3 years.

 

[DrMrLordX]

. Oh, and Nvidia is the incumbent with CUDA.

Personally I think the issue of datacenter compute sales are a bit off-topic since we are talking about AMD's future graphics offerings. Their compute offerings aren't all that great at rendering. They're mostly aimed at fp64 compute and anything else above 8/16-bit precision (nVidia dominates there when their Tensor cores can be used).

The only relevance of that particular topic to the subject at hand is that compute will drive much of the profits for RTG over the next few years. Expect to see more Vega and more GCN, possibly with compute units to compete with nVidia's Tensor cores.

I don't think AI/ML will end up much for Intel.

Watch Loihi. See how it and its successors pan out.

If they won the next gen consoles, I think they'll be fine. Hopefully for Navi they cooked up some cool new features that all the next gen games use.

And there's the actual future of graphics under RTG over the next 3-5 years. AMD will retreat to their base in consoles. Whatever they develop for the consoles will be iterated upon and improved for the PC market in various forms. Radeon VII showed us that AMD is not all that serious about competing on the high-end of the PC desktop. Given how revenue outlooks are for both AMD and nVidia, perhaps AMD made the right choice.

they'll do fine now with a weak GPU.

I'm not sure what appears in the next gen of consoles can be considered "weak". Assuming it's Navi10 or some variation thereof, we should expect to see something low power that performs around Vega56->Vega64 levels, without the encumbrance of non-functional features and other cruft that comes along with Vega.

I think it's more likely that Intel will steal AMD's console business than Nvidia.

Why? Intel likes high margins. AMD can always retreat to lower margins that Intel shareholders would find unacceptable. Plus for the time being, Intel has nothing to show us but Gen11 iGPUs that would be underpowered compared to Navi. Xe isn't going to be ready until 2020 at the earliest, and we don't know exactly when in 2020. Navi should be here in dGPU form by Q4 2019, and console variants will probably be ready for testing earlier than that.

On the other hand we know that AMD Compute GPUs made a record revenue in Q4 2018.

AMD's GPGPU sales have been increasing since they launched mi25. People seem to forget this fact. Vega20 has only helped AMD in that department. But again, we're veering away from the main topic.

Yeap, right now AMD wants to clear all RX 5xx and Vega 56/64 stock and then bring Navi to compete against Turing.

Is Navi really going to compete against Turing? Maybe the 1660Ti? I don't see it as a competitor to the 2080Ti.

12.5% share is pretty much what they had with Bulldozer derivatives in the CPU markets. Yea they have presence, but its really not much.

Maybe at the beginning of the Bulldozer era. But not by the end. It's important to note that, as soon as AMD went with BD-based Opterons, that they lost market share steadily. AMD has gained enterprise dGPU market share under Vega/Vega20.

by the time Gen 11 is here AMD will be here with their next gen.

I don't think that's true. We'll see Gen11 graphics in laptops this June. AMD may not have Navi in any of their PC APUs until 2020. It sure won't be in Picasso.